How to respond immediately?

Question

Hello. I am implementing some chained Servlets to move and manipulate data internally -- there will be no HTML in the response, no presentation of any sort. I am using Servlets only to take advantage of the Server's built-in queuing, scaling and multithreading capabilities, but the protocol here doesn't matter. What's nice is that, if my Servlet isn't ready to handle a request just yet, the Server will hang on to that request for a while. Keeps me from needing a queue. What I don't want, however, is for the client to have to wait until the request has been handled.
Each Servlet does some work and then forwards the request and response on to the next Servlet in the chain. My question is, how can I ensure that as soon as the request is submitted, the client can go on about his business? Don't I have to somehow force the response to be sent back in the first line of code?
When I invoke my first Servlet with Apache Bench, I don't get my prompt back until all the Servlets in the chain have completed their tasks. I realize I can just add "&" to the end of the command-line statement, but that doesn't really solve my problem. I want the Servlet to notify the client immediately that the request was accepted.
Sorry if this is unclear. Hope someone can help.
-Erik

Answer 1

Well, you're a hell of a guy for sticking with me here, but, alas, I tried that, too.
Here is my doPost() in case I am doing something obviously wrong:
public void doPost(HttpServletRequest req, HttpServletResponse res) throws ServletException, IOException {
     System.out.println("Entering doPost in Publisher2");
     String filename = req.getPathInfo();
     try {
     PrintWriter out = res.getWriter();
     out.flush();
     out.close();
     publish(getXMLAsString(filename));
     catch (Exception e) {
     System.out.println(e);
The browser waits until publish() is finished. Tried flushBuffer(), too, as we discussed. I guess the publish() method could spawn a new Thread if there is no other way, but this doesn't seem right somehow. It could be though -- I guess the Thread that's running through doPost() would die just about right after the new Thread was spawned, you know?